SOME EXPERIMENTS ON THE RECOGNITION OF SPEECH, WITH ONE AND WITH 2 EARS
SOME EXPERIMENTS ON THE RECOGNITION OF SPEECH, WITH ONE AND WITH 2 EARS
复制标题
DOI：
10.1121/1.1907229
复制
发表时间：
1953-01-01
影响因子：
2.4
通讯作者：
CHERRY, EC
中科院分区：
文献类型：
--
作者：
CHERRY, EC
同一个人在磁带录音的同一部分上记录的两个不同的信息可以按主题分开。当两种不同的语音信息同时呈现给两只耳朵时，受试者同样可以通过一次只听一个来轻松区分它们。受试者无法识别“被拒绝”消息的单词，甚至语言，但是，如果同一消息用于两只耳朵，并且间隔时间越来越短，受试者在两个消息在耳朵中重合后的2-6秒内识别出相似性。对于由一连串陈词滥调组成的信息，受试者将双耳印象重新组合成由大约。两者的等分。受试者还可以重建随机交替呈现给两只耳朵的信息。这些结果表明，语音识别需要大脑拥有大量的概率排序存储，以便与噪音或干扰作斗争，并做出最大似然估计。
Two different messages recorded by the same person on the same section of a tape recording can be separated by subjects. When 2 different spoken messages are presented simultaneously to the 2 ears, subjects can likewise easily separate them by listening to only one at a time. The words, or even the language, of the "rejected" message are unidentifiable to the subjects, but, if the same message is used for the 2 ears with a decreasing time-lag between, the subjects recognize the similarity 2-6 sec. after the 2 messages coincide in the ears. With messages consisting of strings of cliches, the subjects recombine the binaural impressions into new messages consisting of approx. equal parts of the two. The subjects can also reconstruct messages presented alternately at random to the 2 ears. These results suggest that recognition of speech requires the brain to have a vast store of probability-rankings enabling noise or disturbances to be combatted and maximum-likelihood estimates to be made.